The Application Exchange (APEX) Access Service
The service that decides who may send what to whom across an APEX relaying mesh.
Rose, M., Klyne, G., and D. Crocker, "The Application Exchange (APEX) Access Service", RFC 3341, DOI 10.17487/RFC3341, July 2002, <https://www.rfc-editor.org/info/rfc3341>.
@misc{rfc3341,
series = {Request for Comments},
number = 3341,
howpublished = {RFC 3341},
publisher = {RFC Editor},
doi = {10.17487/RFC3341},
url = {https://www.rfc-editor.org/info/rfc3341},
author = {M. Rose and G. Klyne and D. Crocker},
title = {{The Application Exchange (APEX) Access Service}},
pagetotal = 26,
year = 2002,
month = jul,
} RFC text
Contents
Network Working Group M. Rose
Request for Comments: 3341 Dover Beach Consulting, Inc.
Category: Standards Track G. Klyne
Clearswift Corporation
D. Crocker
Brandenburg InternetWorking
July 2002
The Application Exchange (APEX) Access Service
Status of this Memo
This document specifies an Internet standards track protocol for the
Internet community, and requests discussion and suggestions for
improvements. Please refer to the current edition of the "Internet
Official Protocol Standards" (STD 1) for the standardization state
and status of this protocol. Distribution of this memo is unlimited.
Copyright Notice
Copyright (C) The Internet Society (2002). All Rights Reserved.
Abstract
This memo describes the Application Exchange (APEX) access service,
addressed as the well-known endpoint "apex=access". The access
service is used to control use of both the APEX "relaying mesh" and
other APEX services.
Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 2
2. Use and Management of Access Information . . . . . . . . . . . 3
2.1 Querying Access Information . . . . . . . . . . . . . . . . . 3
2.2 Retrieval of Access Information . . . . . . . . . . . . . . . 4
2.3 Update of Access Information . . . . . . . . . . . . . . . . . 5
3. Format of Access Entries . . . . . . . . . . . . . . . . . . . 9
3.1 Finding the Appropriate Entry: Matching Owners and Actors . . 11
3.2 Creating and Updating Access Entries . . . . . . . . . . . . . 14
4. The Access Service . . . . . . . . . . . . . . . . . . . . . . 14
4.1 Use of XML and MIME . . . . . . . . . . . . . . . . . . . . . 15
4.2 The Query Operation . . . . . . . . . . . . . . . . . . . . . 16
4.3 The Get Operation . . . . . . . . . . . . . . . . . . . . . . 17
4.4 The Set Operation . . . . . . . . . . . . . . . . . . . . . . 18
4.5 The Reply Operation . . . . . . . . . . . . . . . . . . . . . 20
5. Registration: The Access Service . . . . . . . . . . . . . . . 20
6. The Access Service DTD . . . . . . . . . . . . . . . . . . . . 21
7. Security Considerations . . . . . . . . . . . . . . . . . . . 23
References . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . 24
A. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 25
Full Copyright Statement . . . . . . . . . . . . . . . . . . . 26
1. Introduction
This memo describes an access service that is built upon the APEX [1]
"relaying mesh". The APEX access service is used to control use of
both the relaying mesh and other APEX services.
APEX, at its core, provides a best-effort datagram service. Within
an administrative domain, all relays must be able to handle messages
for any endpoint within that domain. APEX services are logically
defined as endpoints but given their ubiquitous semantics they do not
necessarily need to be associated with a single physical endpoint.
As such, they may be provisioned co-resident with each relay within
an administrative domain, even though they are logically provided on
top of the relaying mesh, i.e.,
+----------+ +----------+ +----------+ +---------+
| APEX | | APEX | | APEX | | |
| access | | presence | | report | | ... |
| service | | service | | service | | |
+----------+ +----------+ +----------+ +---------+
| | | |
| | | |
+----------------------------------------------------------------+
| |
| APEX core |
| |
+----------------------------------------------------------------+
That is, applications communicate with an APEX service by exchanging
data with a "well-known endpoint" (WKE).
APEX applications communicate with the access service by exchanging
data with the well-known endpoint "apex=access" in the corresponding
administrative domain, e.g., "apex=access@example.com" is the
endpoint associated with the access service in the "example.com"
administrative domain.
Note that within a single administrative domain, the relaying mesh
makes use of the APEX access service in order to determine if an
originator is allowed to transmit data to a recipient (c.f., Step 5.3
of Section 4.4.4.1 of [1]).
2. Use and Management of Access Information
Access information is organized around access entries, each of which
contains:
o an owner: an APEX address with which the entry is associated;
o an actor: an APEX address that is granted permission to perform
some action in the context of the owner;
o a list of actions; and,
o a timestamp indicating when the service last created or modified
the access entry.
The access entry for a given owner controls access to a potentially
large range of different APEX services, such as data delivery, access
control, and presence information. In addition, Section 4.5 of [1]
discusses APEX access policies that govern such activities as peer
authentication, message relaying, and so on.
Management of access information falls into three categories:
o applications may query the access service to see if one or more
actions are allowed;
o applications may retrieve access information associated with an
owner/actor combination; and,
o applications may modify (i.e., create, replace, or delete) access
information associated with an owner/actor combination.
Each is now described in turn.
2.1 Querying Access Information
When an application wants to determine whether one or more actions
are allowed for an owner/actor combination, it sends a "query"
element to the service, e.g.,
+-------+ +-------+
| | -- data -------> | |
| appl. | | relay |
| | <--------- ok -- | |
+-------+ +-------+
C: <data content='#Content'>
<originator identity='fred@example.com' />
<recipient identity='apex=access@example.com' />
<data-content Name='Content'>
<query owner='fred@example.com' transID='1'
actor='barney@example.com'
actions='core:data presence:subscribe' />
</data-content>
</data>
S: <ok />
The service immediately responds with either an allow or deny
operation containing the same transaction-identifier, where "allow"
means that all of the actions listed in the query are permitted,
e.g.,
+-------+ +-------+
| | <------- data -- | |
| relay | |access |
| | -- ok ---------> | svc. |
+-------+ +-------+
C: <data content='#Content'>
<originator identity='apex=access@example.com' />
<recipient identity='fred@example.com' />
<data-content Name='Content'>
<allow transID='1' />
</data-content>
</data>
S: <ok />
or
C: <data content='#Content'>
<originator identity='apex=access@example.com' />
<recipient identity='fred@example.com' />
<data-content Name='Content'>
<deny transID='1' />
</data-content>
</data>
S: <ok />
2.2 Retrieval of Access Information
When an application wants to retrieve the access entry associated
with an owner/actor combination (typically in preparation for
updating that access information), it sends a "get" element to the
service, e.g.,
+-------+ +-------+
| | -- data -------> | |
| appl. | | relay |
| | <--------- ok -- | |
+-------+ +-------+
C: <data content='#Content'>
<originator identity='fred@example.com' />
<recipient identity='apex=access@example.com' />
<data-content Name='Content'>
<get transID='2'
owner='fred@example.com'
actor='*@example.com' />
</data-content>
</data>
S: <ok />
The service immediately responds with a set operation containing the
access entry and the same transaction-identifier, e.g.,
+-------+ +-------+
| | <------- data -- | |
| relay | |access |
| | -- ok ---------> | svc. |
+-------+ +-------+
C: <data content='#Content'>
<originator identity='apex=access@example.com' />
<recipient identity='fred@example.com' />
<data-content Name='Content'>
<set transID='2'>
<access owner='fred@example.com'
actor='*@example.com'
actions='core:data presence:subscribe'
lastUpdate='2000-05-14T13:02:00-08:00' />
</set>
</data-content>
</data>
S: <ok />
2.3 Update of Access Information
When an application wants to create or modify an access entry
associated with an owner/actor combination, it sends a "set" element
to the service containing the new access entry, e.g.,
+-------+ +-------+
| | -- data -------> | |
| appl. | | relay |
| | <--------- ok -- | |
+-------+ +-------+
C: <data content='#Content'>
<originator identity='wilma@example.com' />
<recipient identity='apex=access@example.com' />
<data-content Name='Content'>
<set transID='1'>
<access owner='fred@example.com'
actor='*@example.com'
actions='core:data presence:subscribe'
lastUpdate='2000-05-14T13:02:00-08:00' />
</set>
</data-content>
</data>
S: <ok />
Note that Step 4 of Section 4.4 requires that the "lastUpdate"
attribute of an access entry be supplied in order to update that
entry; accordingly, applications must successfully retrieve an access
entry prior to trying to modify that entry. (Naturally,
administrators should ensure that applications authorized to modify
an access entry are also authorized to retrieve that entry.)
The service immediately responds with a reply operation containing
the same transaction-identifier, e.g.,
+-------+ +-------+
| | <------- data -- | |
| relay | |access |
| | -- ok ---------> | svc. |
+-------+ +-------+
C: <data content='#Content'>
<originator identity='apex=access@example.com' />
<recipient identity='wilma@example.com' />
<data-content Name='Content'>
<reply code='250' transID='1' />
</data-content>
</data>
S: <ok />
Note that Steps 6.2 and 9.2 of Section 4.4 require that the access
service update the "lastUpdate" attribute of an access entry when it
is created or modified.
The service also immediately sends a set operation to the owner
attribute associated with the access entry, e.g.,
+-------+ +-------+
| | <------- data -- | |
| relay | |access |
| | -- ok ---------> | svc. |
+-------+ +-------+
C: <data content='#Content'>
<originator identity='apex=access@example.com' />
<recipient identity='fred@example.com' />
<data-content Name='Content'>
<set transID='1'>
<access owner='fred@example.com'
actor='*@example.com'
actions='core:data presence:subscribe'
lastUpdate='2000-05-14T23:02:00-08:00' />
</set>
</data-content>
</data>
S: <ok />
When an application wants to delete the access entry associated with
an owner/actor combination, it sends a "set" element to the service
omitting the permitted actions, e.g.,
+-------+ +-------+
| | -- data -------> | |
| appl. | | relay |
| | <--------- ok -- | |
+-------+ +-------+
C: <data content='#Content'>
<originator identity='wilma@example.com' />
<recipient identity='apex=access@example.com' />
<data-content Name='Content'>
<set transID='2'>
<access owner='fred@example.com'
actor='*@example.com'
lastUpdate='2000-05-14T13:02:00-08:00' />
</set>
</data-content>
</data>
S: <ok />
The service immediately responds with a reply operation containing
the same transaction-identifier, e.g.,
+-------+ +-------+
| | <------- data -- | |
| relay | |access |
| | -- ok ---------> | svc. |
+-------+ +-------+
C: <data content='#Content'>
<originator identity='apex=access@example.com' />
<recipient identity='wilma@example.com' />
<data-content Name='Content'>
<reply code='250' transID='2' />
</data-content>
</data>
S: <ok />
The service also immediately sends a set operation to the owner
attribute associated with the access entry, e.g.,
+-------+ +-------+
| | <------- data -- | |
| relay | |access |
| | -- ok ---------> | svc. |
+-------+ +-------+
C: <data content='#Content'>
<originator identity='apex=access@example.com' />
<recipient identity='fred@example.com' />
<data-content Name='Content'>
<set transID='2'>
<access owner='fred@example.com'
actor='*@example.com'
lastUpdate='2000-05-14T13:02:00-08:00' />
</set>
</data-content>
</data>
S: <ok />
Because there are no actions associated with this access entry, the
owner knows that the entry has been deleted.
Note that because access control supported limited wildcarding of
actors, deleting an access entry for a particular owner/actor
combination, may modify, rather than remove, permission. Because of
this, a special action, "all:none", is used.
For example, consider these two access entries:
<access owner='fred@example.com'
actor='barney@example.com'
actions='core:data presence:subscribe presence:watch'
lastUpdate='2000-05-14T13:20:00-08:00' />
<access owner='fred@example.com'
actor='*@example.com'
actions='core:data'
lastUpdate='2000-05-14T13:20:00-08:00' />
Deleting the first access entry will not remove all permissions for
for the actor "barney@example.com".
Instead, the first access entry should be modified thusly:
<access owner='fred@example.com'
actor='barney@example.com'
actions='all:none'
lastUpdate='2000-05-14T13:20:00-08:00' />
3. Format of Access Entries
Each administrative domain is responsible for maintaining one or more
"access entries" for each of its endpoints and associated
subaddresses (regardless of whether those addresses are currently
attached to the relaying mesh).
A separate access entry is required for each actor or group of actors
for whom access permission is specified. Section 6 defines the
syntax for access entries. Each access entry has an "owner"
attribute, an "actor" attribute, an "actions" attribute, a
"lastUpdate" attribute, and no content:
o the "owner" attribute specifies the address (endpoint or
subaddress) associated with the access entry;
o the "actor" attribute specifies an entity or group of entities for
whom access permissions are specified, as described below;
o the "actions" attribute specifies the permissions granted to the
actor in the context of the owner; and,
o the "lastUpdate" attribute specifies the date and time that the
service last created or modified the access entry.
An action is specified as a service/operation pair, e.g., the action
"presence:publish" refers to the "publish" operation of the
"presence" service. Two service values are reserved:
o "all" is used to refer to all services, e.g., "all:data"; and,
o "core" is used to refer to the service implemented by the relaying
mesh, e.g., the "core:data" permission is consulted by the
relaying mesh (c.f., Step 5.3 of Section 4.4.4.1 of [1]).
Further, two operation values are reserved:
o "all" is used to refer to all operations, e.g., "presence:all";
and,
o "none" is used to refer to no operations whatsoever, e.g.,
"all:none".
An actor is an APEX address and is specified using the "entity"
syntax specified in Section 2.2 of [1]. However, both the "local"
and "domain" parts may contain limited wildcarding:
o The "local" part is either:
* a literal string (e.g., "fred");
* a subaddress wildcard (e.g., "fred/*" or "apex=pubsub/*"); or,
* the value "apex=*", specifying all APEX services;
* the value "*", specifying any address other than an APEX
service.
o The "domain" part is either:
* a FQDN (e.g., "example.com");
* a domain wildcard (e.g., "*.example.com"); or,
* the value "*", specifying all administrative domains.
Note that in the case of a domain wildcard, the wildcard itself
matches zero or more subdomains, e.g., "*.example.com" matches
"example.com", "foo.example.com", "bar.foo.example.com", and so
on.)
The following default entries are provided for each owner, but are
overridden by an explicitly supplied entry with the same actor value:
actor='local@domain' actions='all:all'
actor='apex=*@domain' actions='all:all'
actor='apex=*@*' actions='core:data'
actor='*@*' actions='all:none'
where "local@domain" specifies the owner associated with the access
entry.
For example, the explicit entry
actor='*@*' actions='core:data'
allows endpoints from any domain to use the relaying mesh to send
data to the owner, but does not override the default entry for
"apex=*@domain", which allows all APEX services in the owner's domain
access to all actions.
APEX endpoint names can legitimately contain the character '*', but
access entries use '*' to indicate wildcarding. Accordingly, the
two-character sequence '\*' is used to avoid ambiguity in the "actor"
attribute. Similarly, to explicitly specify an endpoint name
containing '\' in the "actor" attribute, the two-character sequence
'\\' is used.
Note that this convention is used only for the "actor" attribute of
the "get" operation and of the "access" entry that appears in the
"set" operation; however, this convention is not used in the "query"
operation, as this operation does not allow wildcarding.
For example, to specify the endpoint named as "a\b*c@example.com" in
the "get" operation or in an "access" entry, the string
"a\\b\*c@example.com" is used; but in the "query" operation, the
string "a\b*c@example.com" is used. (Of course, as name allocation
is a local matter, these complications can be avoided by the simple
expedient of not using endpoint names containing '*' or '\'.)
3.1 Finding the Appropriate Entry: Matching Owners and Actors
The use of actor wildcarding makes it possible for several access
entries to apply for a given owner/actor combination. When
determining which access entry to use when responding to the query
operation, the algorithm is:
o Consider only those access entries that are associated with the
given owner.
o Consider only those access entries in which the actor value
matches the actor address in the query. If the wildcard character
('*') is present, then it a match is possible only if each
wildcard character can be replaced with a non-empty character
sequence (one or more characters) to obtain a value identical to
the address in the query.
o Order those remaining access entries:
* Use the exactness of the match with the domain part of the
actor value as the primary key; and,
* Use the exactness of the match with the local part of the actor
value as the secondary key.
o When matching with the domain part, an exact match is the best
match; otherwise, the shorter the wildcard match, the higher the
priority.
For example, if the actor's domain is "bar.foo.example.com", a
match against an entry of "*.foo.example.com" is better than a
match against an entry of "*.example.com".
o When matching with the local part, an exact match is the best
match; otherwise, the shorter the wildcard match, the higher the
priority. This is true regardless of whether the wildcarding is
for subaddress or service. (Note that a local part with a
wildcard subaddress does not have a non-empty match with the same
local part without a subaddress.)
For example, consider these access entries:
<access owner='fred@example.com'
actor='wilma@example.com'
actions='all:all'
lastUpdate='2000-05-14T13:20:00-08:00' />
<access owner='fred@example.com'
actor='mr.slate@example.com'
actions='core:data'
lastUpdate='2000-05-14T13:20:00-08:00' />
<access owner='fred/appl=wb@example.com'
actor='barney/appl=wb@example.com'
actions='core:data'
lastUpdate='2000-05-14T13:20:00-08:00' />
<access owner='fred@example.com'
actor='*@example.com'
actions='core:data presence:subscribe presence:watch'
lastUpdate='2000-05-14T13:20:00-08:00' />
<access owner='fred@example.com'
actor='*@*'
actions='core:data'
lastUpdate='2000-05-14T13:20:00-08:00' />
Briefly:
o For addresses within the "example.com" administrative domain:
* "fred", "wilma", and all APEX services within the "example.com"
administrative domain are allowed access to all operations for
"fred@example.com";
* "mr.slate" is allowed access only to send data through the
relaying mesh to "fred@example.com";
* "barney/appl=wb" is allowed access only to send data to "fred/
appl=wb", a subaddress of "fred@example.com"; and,
* any other address within the "example.com" administrative
domain is allowed access to send data and invoke the
"subscribe" and "watch" operations of the APEX presence service
with respect to "fred@example.com".
o For any address outside the "example.com" administrative domain,
the address is allowed access to send data, regardless of whether
it is an APEX service.
Note that although the four default entries are always available, the
explicit entry for actor "*@*" overrides the corresponding default
entry.
3.2 Creating and Updating Access Entries
The get and set operations are provided as a basic mechanism for
creating and updating access rules, for which no special wildcard
processing is performed.
The actor value for an access entry may contain limited wildcard
characters which have special significance only when performing a
query operation (cf., Section 3.1). For the purposes of retrieving
and updating entries, actor values are treated simply as literal
names.
4. The Access Service
Section 5 contains the APEX service registration for the access
service:
o Within an administrative domain, the service is addressed using
the well-known endpoint of "apex=access".
o Section 6 defines the syntax of the operations exchanged with the
service.
o A consumer of the service initiates communications by sending data
containing a query, get, or set operation.
o The service replies to these operations.
o When an access entry is changed, the service sends a notification
to the owner associated with the changed entry.
An implementation of the service must maintain information about
access entries in persistent storage.
Consult Section 6.1.1 of [1] for a discussion on the properties of
long-lived transaction-identifiers.
4.1 Use of XML and MIME
Section 4.1 of [1] describes how arbitrary MIME content is exchanged
as a BEEP [2] payload. For example, to transmit:
<data content='...'>
<originator identity='fred@example.com' />
<recipient identity='apex=access@example.com' />
</data>
where "..." refers to:
<query owner='fred@example.com' transID='1'
actor='barney@example.com'
actions='core:data presence:subscribe' />
then the corresponding BEEP message might look like this:
C: MSG 1 2 . 42 1234
C: Content-Type: multipart/related; boundary="boundary";
C: start="<1@example.com>";
C: type="application/beep+xml"
C:
C: --boundary
C: Content-Type: application/beep+xml
C: Content-ID: <1@example.com>
C:
C: <data content='cid:2@example.com'>
C: <originator identity='fred@example.com' />
C: <recipient identity='apex=access@example.com' />
C: </data>
C: --boundary
C: Content-Type: application/beep+xml
C: Content-ID: <2@example.com>
C:
C: <query owner='fred@example.com' transID='1'
C: actor='barney@example.com'
C: actions='core:data presence:subscribe' />
C: --boundary--
C: END
or this:
C: MSG 1 1 . 42 267
C: Content-Type: application/beep+xml
C:
C: <data content='#Content'>
C: <originator identity='fred@example.com' />
C: <recipient identity='apex=access@example.com' />
C: <data-content Name='Content'>
C: <query owner='fred@example.com' transID='1'
C: actor='barney@example.com'
C: actions='core:data presence:subscribe' />
C: </data-content>
C: </data>
C: END
4.2 The Query Operation
When an application wants to see if a particular operation is
allowed, it sends a "query" element to the service.
The "query" element has an "owner" attribute, an "actor" attribute,
an "actions" attribute, a "transID" attribute, and no content:
o the "owner" attribute specifies the address associated with the
access entry;
o the "actor" attribute specifies the address (without wildcarding)
for which access permissions are queried;
o the "actions" attribute specifies one or more actions for which
permission is queried; and,
o the "transID" attribute specifies the transaction-identifier
associated with this operation.
When the service receives a "query" element, we refer to the "owner"
attribute as the "subject". The service performs these steps:
1. If the subject is outside this administrative domain, a "reply"
element having code 553 is sent to the originator.
2. If the subject does not refer to a valid address, a "reply"
element having code 550 is sent to the originator.
3. If the subject's access entry matching the originator does not
contain an "access:query" token, a "reply" element having code 537
is sent to the originator.
4. The subject's access entry matching the actor attribute of the
query element is selected (cf., Section 3.1).
5. If all of the permissions in the "actions" attribute of the query
element are contained in the selected access entry, then an
"allow" element is sent to the originator.
6. Otherwise, a "deny" element is sent to the originator.
Regardless of whether an "allow", "deny", or "reply" element is sent
to the originator, the "transID" attribute is identical to the value
found in the "query" element sent by the originator.
4.3 The Get Operation
Prior to creating or updating an access entry for some owner/actor
combination, an application will usually need to retrieve any
existing access entry. It does so by sending a "get" element to the
service. In particular, a successful response returns a "lastUpdate"
value that is necessary when sending a subsequent "set" element.
The "get" element has an "owner" attribute, an "actor" attribute, a
"transID" attribute, and no content:
o the "owner" attribute specifies the address associated with the
access entry;
o the "actor" attribute specifies the address (with possible
wildcarding) for which access permissions are retrieved; and,
o the "transID" attribute specifies the transaction-identifier
associated with this operation.
When the service receives a "get" element, we refer to the "owner"
attribute as the "subject". The service performs these steps:
1. If the subject is outside this administrative domain, a "reply"
element having code 553 is sent to the originator.
2. If the subject does not refer to a valid address, a "reply"
element having code 550 is sent to the originator.
3. If the subject's access entry matching the originator does not
contain an "access:get" token, a "reply" element having code 537
is sent to the originator.
4. The subject's access entry whose "actor" attribute identically
matches the "actor" attribute of the "get" element is selected.
5. If no such entry exists, a "reply" element having code 551 is sent
to the originator.
6. Otherwise, a "set" element corresponding to the selected access
entry is sent to the originator.
Regardless of whether a "set" or "reply" element is sent to the
originator, the "transID" attribute is identical to the value found
in the "get" element sent by the originator.
4.4 The Set Operation
When an application wants to modify (i.e., create, replace, or
delete) the access entry associated with an owner/actor combination,
it sends a "set" element to the service.
The "set" element has a "transID" attribute, and contains an "access"
element:
o the "transID" attribute specifies the transaction-identifier
associated with this operation; and,
o the "access" element contains the access entry to be created,
replaced, or deleted.
The "access" element has an "owner" attribute, an "actor" attribute,
an optional "actions" attribute, an optional "lastUpdate" attribute,
and no content:
o the "owner" attribute specifies the address associated with the
access entry;
o the "actor" attribute specifies the address (with possible
wildcarding) for which access permissions are specified;
o the "actions" attribute (present only to add or replace an entry)
specifies one or more actions for which permission is to be
determined; and,
o the "lastUpdate" attribute (present only to replace or delete an
entry) specifies the current timestamp of the access entry that is
to be replaced.
When the service receives a "set" element, we refer to the "owner"
attribute of the access element as the "subject". The service
performs these steps:
1. If the subject is outside this administrative domain, a "reply"
element having code 553 is sent to the originator.
2. If the subject does not refer to a valid address, a "reply"
element having code 550 is sent to the originator.
3. If the subject's access entry matching the originator does not
contain an "access:set" token, a "reply" element having code 537
is sent to the originator.
4. The subject's access entry whose "actor" attribute identically
matches the "actor" attribute of the "set" element is selected.
5. If no such entry exists and the "lastUpdate" attribute is present
in the supplied "set" element, a "reply" element having code 555
is sent to the originator.
6. If no such entry exists and the "lastUpdate" attribute is absent
in the supplied "set" element, then:
1. The access entry for the owner/actor combination is created
from the supplied "access" element.
2. The "lastUpdate" attribute of that access entry set to the
service's notion of the current date and time.
3. A "reply" element having code 250 is sent to the originator.
4. A "set" element corresponding to the newly-created access entry
is sent to the subject's address.
7. If the selected entry exists, but its "lastUpdate" attribute is
not semantically identical to the "lastUpdate" attribute of the
supplied "access" element, a "reply" element having code 555 is
sent to the originator.
8. If "actions" attribute of the supplied "access" element is not
present, then:
1. The selected entry is deleted.
2. A "reply" element having code 250 is sent to the originator.
3. A "set" element corresponding to the owner/actor combination,
but lacking an "actions" attribute is sent to the subject's
address.
9. Otherwise:
1. The access entry for the owner/actor combination is updated
from the supplied "access" element.
2. The "lastUpdate" attribute of the updated access entry is set
to the service's notion of the current date and time (which
should be different from the "lastUpdate" value associated with
any replaced entry).
3. A "reply" element having code 250 is sent to the originator.
4. A "set" element corresponding to the newly-updated access entry
is sent to the subject's address.
When sending the "reply" element, the "transID" attribute is
identical to the value found in the "set" element sent by the
originator.
4.5 The Reply Operation
While processing operations, the service may respond with a "reply"
element. Consult Sections 10.2 and 6.1.2 of [1], respectively, for
the definition and an exposition of the syntax of the reply element.
5. Registration: The Access Service
Well-Known Endpoint: apex=access
Syntax of Messages Exchanged: c.f., Section 6
Sequence of Messages Exchanged: c.f., Section 4
Access Control Tokens: access:query, access:get, access:set
Contact Information: c.f., the "Authors' Addresses" section of this
memo
6. The Access Service DTD
<!--
DTD for the APEX access service, as of 2001-06-19
Refer to this DTD as:
<!ENTITY % APEXACCESS PUBLIC "-//IETF//DTD APEX ACCESS//EN" "">
%APEXACCESS;
-->
<!ENTITY % APEXCORE PUBLIC "-//IETF//DTD APEX CORE//EN" "">
%APEXCORE;
<!--
DTD data types:
entity syntax/reference example
====== ================ =======
access actor
ACTOR an ENDPOINT or a *@example.com
wildcard
permitted actions
ACTIONS a list of access "core:any access:query"
tokens
-->
<!ENTITY % ACTOR "CDATA">
<!ENTITY % ACTIONS "NMTOKENS">
<!--
Synopsis of the APEX access service
service WKE: apex=access
message exchanges:
consumer initiates service replies
================== ================
query allow, deny, or reply
get set or reply
set reply
service initiates consumer replies
================= ================
set (nothing)
access control:
token target
========== ======
access:query for "owner" of "access" element
access:get for "owner" of "access" element
access:set for "owner" of "access" element
-->
<!ELEMENT query EMPTY>
<!ATTLIST query
owner %ENDPOINT; #REQUIRED
actor %ACTOR; #REQUIRED
actions %ACTIONS; #REQUIRED
transID %UNIQID; #REQUIRED>
<!ELEMENT get EMPTY>
<!ATTLIST get
owner %ENDPOINT; #REQUIRED
actor %ACTOR; #REQUIRED
transID %UNIQID; #REQUIRED>
<!ELEMENT set (access)>
<!ATTLIST set
transID %UNIQID; #REQUIRED>
<!ELEMENT allow EMPTY>
<!ATTLIST allow
transID %UNIQID; #REQUIRED>
<!ELEMENT deny EMPTY>
<!ATTLIST deny
transID %UNIQID; #REQUIRED>
<!--
access entries
-->
<!ELEMENT access EMPTY>
<!ATTLIST access
owner %ENDPOINT; #REQUIRED
actor %ACTOR; #REQUIRED
actions %ACTIONS; #IMPLIED
lastUpdate %TIMESTAMP; #IMPLIED>
7. Security Considerations
Consult [1]'s Section 11 for a discussion of security issues.
In addition, timestamps issued by the the access service may disclose
location information. If this information is considered sensitive,
the special timezone value "-00:00" may be used (after converting the
local time accordingly).
References
[1] Rose, M., Klyne, G. and D. Crocker, "The Application Exchange
Core", RFC 3340, July 2002.
[2] Rose, M., "The Blocks Extensible Exchange Protocol Core", RFC
3080, March 2001.
Authors' Addresses
Marshall T. Rose
Dover Beach Consulting, Inc.
POB 255268
Sacramento, CA 95865-5268
US
Phone: +1 916 483 8878
EMail: mrose@dbc.mtview.ca.us
Graham Klyne
Clearswift Corporation
1310 Waterside
Arlington Business Park
Theale, Reading RG7 4SA
UK
Phone: +44 11 8903 8903
EMail: Graham.Klyne@MIMEsweeper.com
David H. Crocker
Brandenburg Consulting
675 Spruce Drive
Sunnyvale, CA 94086
US
Phone: +1 408 246 8253
EMail: dcrocker@brandenburg.com
URI: http://www.brandenburg.com/
Appendix A. Acknowledgements
The authors gratefully acknowledge the contributions of: Neil Cook,
Darren New, Chris Newman, Scott Pead, and Bob Wyman.
Full Copyright Statement
Copyright (C) The Internet Society (2002). All Rights Reserved.
This document and translations of it may be copied and furnished to
others, and derivative works that comment on or otherwise explain it
or assist in its implementation may be prepared, copied, published
and distributed, in whole or in part, without restriction of any
kind, provided that the above copyright notice and this paragraph are
included on all such copies and derivative works. However, this
document itself may not be modified in any way, such as by removing
the copyright notice or references to the Internet Society or other
Internet organizations, except as needed for the purpose of
developing Internet standards in which case the procedures for
copyrights defined in the Internet Standards process must be
followed, or as required to translate it into languages other than
English.
The limited permissions granted above are perpetual and will not be
revoked by the Internet Society or its successors or assigns.
This document and the information contained herein is provided on an
"AS IS" basis and THE INTERNET SOCIETY AND THE INTERNET ENGINEERING
TASK FORCE DISCLAIMS ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING
BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION
HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF
MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
Acknowledgement
Funding for the RFC Editor function is currently provided by the
Internet Society.
ISO Plain Language
On this page
§1What the APEX Access Service Does
The APEX access service is built on top of the APEX relaying mesh and is used to control use of that mesh and of the other APEX services. APEX, at its core, provides a best-effort datagram service. Within an administrative domain, every relay must be able to handle messages for any endpoint in that domain.
APEX services are logically defined as endpoints, but because they have ubiquitous semantics they do not need to sit on a single physical endpoint. They may instead be provisioned alongside each relay in the domain, even though they are logically provided on top of the relaying mesh. Applications reach a service by exchanging data with a well-known endpoint. The access service lives at the well-known endpoint apex=access, so apex=access@example.com is the access service in the example.com administrative domain.
The access service is not only there for applications. Within a single administrative domain the relaying mesh itself consults it, to decide whether an originator is allowed to transmit data to a recipient.
APEX relays can carry a message to any address inside the organisation, so something has to check who is allowed to send to whom. That something is the access service, which every relay in the domain can ask.
§3Access Entries and Wildcards
Each administrative domain maintains one or more access entries for each of its endpoints and associated subaddresses, whether or not those addresses are currently attached to the relaying mesh. A separate entry is needed for every actor or group of actors whose permissions are specified. Each entry has four attributes and no content: owner, the address the entry is about; actor, the address granted permission; actions, the permissions granted; and lastUpdate, when the service last created or changed the entry.
An action is a service and operation pair, so presence:publish means the publish operation of the presence service. Two service names are reserved: all, meaning every service, and core, meaning the service implemented by the relaying mesh itself, so that the core:data permission is the one the mesh consults. Two operation names are reserved: all, meaning every operation, and none, meaning no operation at all, written all:none.
The actor is an APEX address, and both its local part and its domain part may carry limited wildcards. The local part may be a literal name such as fred, a subaddress wildcard such as fred/* or apex=pubsub/*, the value apex=* meaning all APEX services, or a lone asterisk meaning any address other than an APEX service. The domain part may be a fully qualified name, a domain wildcard such as *.example.com, or a lone asterisk meaning all administrative domains. A domain wildcard matches zero or more subdomains, so *.example.com matches example.com itself as well as foo.example.com and bar.foo.example.com.
Four default entries are provided for each owner, and an explicitly supplied entry with the same actor value overrides the matching default:
local@domaingetsall:all, that is, every operation of every service;apex=*@domaingetsall:all;apex=*@*getscore:data;*@*getsall:none.
So an explicit entry giving *@* the actions core:data lets endpoints in any domain send data to the owner through the relaying mesh, without disturbing the default that gives the owner's own domain full access. APEX endpoint names may legitimately contain an asterisk, so access entries use a backslash followed by an asterisk to mean a literal one, and two backslashes for a literal backslash. That convention applies to the actor attribute of the get operation and of the access entry in a set operation, but not to the query operation, which allows no wildcards at all.
Each permission record names one address (the owner), one address or pattern (the actor), and what the actor may do to the owner. Patterns let one record cover a whole group, such as everyone in a domain. A record that names one person beats a pattern, and a pattern that is more specific beats a broader one.
| Actor | Default actions |
|---|---|
| local@domain | all:all |
| apex=*@domain | all:all |
| apex=*@* | core:data |
| *@* | all:none |
§3.1Choosing the Right Access Entry
Because actors may carry wildcards, several access entries can apply to a given owner and actor. The RFC gives the algorithm for choosing between them. First, consider only the entries associated with the given owner. Then consider only those entries whose actor value matches the actor address in the query, where a wildcard character can be replaced only by a non-empty sequence of characters to make the match.
The remaining entries are ordered with two keys. The primary key is how exact the match is on the domain part of the actor value, and the secondary key is how exact the match is on the local part. On the domain side an exact match is best, and otherwise the shorter the wildcard match the higher the priority, so for an actor whose domain is bar.foo.example.com a match against *.foo.example.com is better than one against *.example.com. The local part works the same way, whether the wildcard is a subaddress or a service, with one extra note: a local part with a wildcard subaddress does not match the same local part without one.
The RFC works this through with a set of entries for the owner fred@example.com. An entry naming wilma@example.com with all operations covers wilma and the APEX services of the domain for every operation. An entry naming mr.slate@example.com with core:data lets mr.slate send data through the mesh but nothing else. An entry whose owner is the subaddress fred/appl=wb and whose actor is barney/appl=wb lets that actor send data to that subaddress only. An entry for *@example.com with core:data, presence:subscribe and presence:watch gives the rest of the domain those three permissions. An entry for *@* with core:data gives any address outside the domain the right to send data, whether or not it is an APEX service. Although the four default entries are always available, an explicit entry for *@* overrides the matching default.
When more than one record could apply, the service picks the tightest one. An exact address beats a pattern, and among patterns the shorter one wins, so a rule for one subdomain is preferred over a rule for the whole domain.
§4The Three Operations
A consumer of the service starts by sending data containing a query, get or set operation, and the service replies. When an access entry changes, the service sends a notification to the owner. An implementation must keep access entries in persistent storage.
The query operation asks whether one or more actions are allowed. It carries an owner, an actor without wildcards, a list of actions and a transaction identifier, and it has no content. The service treats the owner as the subject and then: sends a reply with code 553 if the subject is outside this administrative domain; sends a reply with code 550 if the subject is not a valid address; sends a reply with code 537 if the subject's entry matching the originator does not contain an access:query token; selects the entry matching the actor as described above; then sends an allow element if all the permissions in the query are contained in the selected entry, and a deny element otherwise. The transaction identifier in the reply is always the one from the query.
The get operation retrieves the entry for an owner and actor, usually so that it can be updated afterwards, because a successful get returns the lastUpdate value that a later set needs. It carries an owner, an actor, which may contain wildcards, and a transaction identifier. The service again answers 553, 550 or 537, the last now for an access:get token, then selects the entry whose actor attribute matches the requested actor exactly. If there is no such entry it replies with code 551, and otherwise it returns a set element carrying that entry.
The set operation creates, replaces or deletes an entry. It carries a transaction identifier and an access element, which has an owner, an actor, an optional actions attribute and an optional lastUpdate attribute. The actions attribute is present only when adding or replacing an entry, and the lastUpdate attribute only when replacing or deleting one, in which case it gives the current timestamp of the entry to be replaced. The service answers 553, 550 or 537, the last now for an access:set token, and selects the entry whose actor matches exactly. If no entry exists but a lastUpdate was supplied, it replies with code 555. If no entry exists and no lastUpdate was supplied, it creates the entry, sets its lastUpdate to the service's own current date and time, replies with code 250 and sends a set element for the new entry to the subject. If the entry exists but its lastUpdate differs from the one supplied, it replies 555. If the supplied access element has no actions attribute, the entry is deleted, a reply with code 250 is sent, and a set element without actions goes to the subject. Otherwise the entry is updated, its lastUpdate is set to the service's current date and time, a reply with code 250 is sent, and a set element for the updated entry goes to the subject.
There are three things a client can do. It can ask a yes or no question about a permission, read back a permission record, or write a new one. Reading and writing use the exact actor name as a key, so wildcards only matter when asking the question.
<data content='#Content'>
<originator identity='fred@example.com' />
<recipient identity='apex=access@example.com' />
<data-content Name='Content'>
<query owner='fred@example.com' transID='1'
actor='barney@example.com'
actions='core:data presence:subscribe' />
</data-content>
</data>
<data content='#Content'>
<originator identity='apex=access@example.com' />
<recipient identity='fred@example.com' />
<data-content Name='Content'>
<allow transID='1' />
</data-content>
</data>- 250Success — The entry was created, updated or deleted.
- 537Permission denied — The originator's own entry lacks the access:query, access:get or access:set token.
- 550Invalid address — The subject does not refer to a valid address.
- 551Entry not found — A get found no access entry with that exact actor value.
- 553Outside the domain — The subject is outside this administrative domain.
- 555Timestamp mismatch — The supplied lastUpdate does not match the stored entry, or an entry was to be replaced but none exists.
§4.1How the Messages Are Carried
The operations are XML elements, carried as MIME content exchanged as a BEEP payload. The RFC shows two ways to do this. The first uses a multipart/related message whose parts have the type application/beep+xml, with the outer data element referring to the second part by a content identifier. The second puts everything in a single application/beep+xml message, with the data element carrying a data-content element named Content that holds the operation. The RFC also prints the document type definition for the service, which declares the query, get, set, allow and deny elements with their attributes, and the access element with its owner, actor, optional actions and optional lastUpdate attributes.
One point about deleting needs care. Because access control supports limited wildcarding of actors, deleting the entry for a particular owner and actor combination may modify permission rather than remove it. The RFC therefore uses the special action all:none for that case: an entry for barney@example.com listing core:data, presence:subscribe and presence:watch is not replaced by a deletion, because the separate entry for *@example.com would still grant some of those permissions. Setting that entry's actions to all:none is the way to express the intent to remove them.
The RFC is also explicit that the get and set operations perform no special wildcard processing. An actor value in an access entry may contain wildcard characters, but those have special significance only when a query is performed; for retrieval and update the actor values are treated simply as literal names.
The messages are XML written in the usual markup style, carried inside MIME containers over BEEP. The RFC also gives the grammar, so two implementations know the same element names mean the same thing.
§7Security Considerations
The RFC refers the reader to the APEX core document for a discussion of security issues, and adds one point of its own. Timestamps issued by the access service may disclose location information, because a timestamp carries a time zone offset. If that information is considered sensitive, the special timezone value -00:00 may be used, after converting the local time accordingly.
The service stamps each permission record with the time it last changed. That stamp includes a time zone offset, which can reveal roughly where the service is, so the RFC offers a way to blank that out.